Conversion of the vocal tract shape for spectral warping by a PARCOR analysis—synthesis system

Abstract

This paper presents a method of spectral warping for helium speech unscrambling by means of converting the vocal tract shape estimated from the PARCOR analysis. Speech produced in a high‐pressure, helium‐oxygen atmosphere has particular distortions. One peculiarity is a disparity in the rates of formant frequency shift and fundamental frequency shift. The formant frequency shift consists of two factors. One is due to the sound velocity change, rendering a linear shift, and the other is due to the high pressure, rendering a nonlinear shift. In order to convert the helium speech to normal speech, therefore, the authors have developed an analysis—synthesis system involving various types of conversion processes for spectral warping. The conversion process presented in this paper includes two operations. The first is to increase the number of sections in an acoustic tube model of the vocal tract for the linear shift, and the second is to modify the vocal tract area function for the nonlinear shift. Experimenta...